Blind Source Separation Under Signal Covariance Constraints: Criteria and Algorithms
S.C. Douglas, Timothy H. DeFries · 2021 55th Asilomar Conference on Signals, Systems, and Computers · 2021
This paper presents criteria and algorithms for blind source separation of linearly-mixed signals under a known set of source covariance constraints through a user-specified spatial covariance matrix. Such techniques are important in practical situations where the sources are expected to have a correlation structure and extend well-known ICA and BSS methods to this novel problem domain. Source dependence is modeled using an approximation to the theory of copulas. The approaches are applied in a real-world remote sensing application whereby exhaust and evaporative plumes from moving vehicles on a roadway are estimated from multiple spectral images collected by a laser-scanning camera.